Twenty-two adult subjects were monitored during sleep and then recorded with eyes open under baseline conditions at a feedback laboratory. Half of these subjects subsequently received three months of bilateral EEG biofeedback training during three 30-minute sessions per week. Light and tone reward was obtained for production of 1 sec trains of either 10, 13 or 15 cps central cortical activity, or 10 cps occipital activity. Central rhythms were rewarded when train voltages exceeded 5 micron v and were less than 60 micron v. Occipital 10 cps was rewarded when train voltages exceeded 10 micron v. All subjects were again monitored for sleep and under baseline conditions after the three month period. Analysis of baseline data indicated that 27 percent of subjects had such low voltage, unmodulated EEG characteristics as to preclude biofeedback training. Of the subjects receiving biofeedback training, 50 percent demonstrated no significant acquisition of the rewarded EEG rhythm, irrespective of the origin of this activity. Subjects showing statistically significant acquisition demonstrated several different patterns. In some, clear acquisition was demonstrated within biofeedback sessions but no increase in absolute level was observed across sessions. Others showed minimal increment within sessions, but a significant increase across sessions. Training of central 13 cps was accompanied by a parallel increment in central 15 cps and a gradual decrease in central 10 cps. specific changes in topographic organization were noted after training. Finally, changes in EEG spindle configuration and sleep stage patterning were differentially associated with the two patterns of acquisition described above. Subjects not demonstrating acquisition showed no significant baseline or slpp parameter changes.